Sparse-sensing-based reduced modeling for rapid characterization of nonlinear magnetic inductors

Rapid prototyping of magnetic inductors requires the evaluation of inductance over a wide range of turn counts and operating currents. This task becomes computationally or experimentally demanding when nonlinear magnetic materials are considered because each candidate design may require iterative solution of the magnetic circuit model or repeated evaluations. This paper proposes a sparse-sensing framework for reconstructing inductance curves using a small number of strategically selected turn counts. A reduced basis is constructed by singular value decomposition (SVD), and informative sensing turn counts are selected by QR decomposition with column pivoting. The resulting model reconstructs the full inductance–turn relationship from only a few evaluations. The method is demonstrated using simulated nonlinear toroidal inductor data and then incorporated into a multi-objective design workflow to select candidate cores from a commercial database. The results show that five selected turn counts are sufficient to reconstruct the inductance curve with a relative error below 0.1% in the validation cases under consideration. Combined with the non-dominated sorting genetic algorithm II (NSGA-II) optimization, the reduced model enables rapid screening of feasible core, material, and stacking configurations while preserving accuracy sufficient for engineering design.